Aixin Yuanzhi was awarded the "Most Innovative Product Award" at the Artificial Intelligence Conference and officially released the Aixin Pi Pro
Time:2023-08-26
Views:725
Aixin Yuanzhi announced that its chip product, Aixin Yuanzhi AX650N, has won the "Most Innovative Product Award" at the 4th Artificial Intelligence Excellence Innovation Award. At the same time, the enterprise has officially released the developer kit - Aixinpai Pro, to facilitate community developers to experience low-cost local deployment of visual models on the edge and end sides, and to create an ecological platform for developers.
On August 23rd, Tang Qi, Director of AI Reasoning Engine at Aixin Yuanzhi, was invited to attend the 7th Artificial Intelligence Conference 2023 jointly organized by Electronic Enthusiasts Network and Elexcon Shenzhen International Electronics Exhibition, and gave a keynote speech on "Deployment of Transformer Visual Model on the Edge Side", sharing the implementation practice of deploying Transformer visual model based on Aixin Tongyuan mixed precision NPU, and officially releasing the developer kit - Aixin Pai Pro. The 4th Artificial Intelligence Excellence Innovation Award was awarded simultaneously. With its strong self-developed strength and outstanding achievements in the edge AI circuit, the Aixin Yuanzhi AX650N was awarded the "Most Innovative Product Award".
Rapid Popularization of End to Side Intelligence Transformer Visual Model Applications to New Heights
Since 2015, the object recognition accuracy of AI models surpassed that of humans for the first time in ImageNet‘s image classification competition, the artificial intelligence industry has officially entered the new era of navigation. In recent years, with the progress of the chip industry process and the trend of miniaturization and lightweight of AI models, the localized deployment of AI models in products such as sweeping robots, home cameras, and smart speakers has become a reality, and end-to-end intelligence has been popularized.
So far, the computing power mainline of cloud, edge, and end triangulation based on traditional CNN models has been completed, "Tang Qi said in a speech. In the process of AI model migration from cloud to end-to-end, the Transformer network structure is also constantly improving, and has attracted attention with the popularity of ChatGPT. Looking back at the development history of Transformer visual models, from the application of semantic segmentation models based on Transformer network structure to autonomous driving scenes, to the segmentation of everything in SAM and the base vision model DINOV2, the application of Transformer visual models is moving towards new heights.
Committed to building a world leading artificial intelligence visual chip, Aixin Yuanzhi focuses on the development of high-performance and low-power edge side and end side artificial intelligence processor chips, laying out three major business markets: smart city, intelligent driving, and AIoT. Based on the upgrading and transformation of smart cities, the large-scale implementation of L2/L2+intelligent driving applications, and the expansion of demand for intelligent terminal devices, Aixin Yuanzhi will continue to invest in edge perception intelligent chips, continuously building sensing and computing infrastructure.
Accelerate the Implementation Efficiency of AI Development by Visually Connecting the Digital and Physical Worlds
From smart cities to intelligent driving to AIoT, the reason why Aixin Yuanzhi can achieve full coverage of the three major application scenarios is based on the company‘s self-developed Aixin Zhimou AI-ISP and Aixin Tongyuan hybrid precision NPU two core technologies.
Aixin Smart Eye AI-ISP combines deep learning algorithms with traditional ISP processing units, serving as the "eye" of all things intelligence to perceive more information, especially for the needs of dark and full color. It has six major technological highlights: AI starlight full color, AI HDR imaging, AI multispectral fusion, AI anti shake, AI scene enhancement, and AI multi-sensor fusion. It can comprehensively improve image quality and achieve better visual effects in multiple driving scenarios such as dark roads, tunnel exits, and bumpy roads.
As another core self-developed technology of Aixin Yuanzhi, Aixin Tongyuan Hybrid Precision NPU supports three calculation accuracies: INT4/INT8/INT16, and has outstanding advantages of high performance, low cost, and easy to use. It is worth mentioning that, drawing on market feedback from the previous two generations of products and combining with the development trend of cutting-edge technology in the industry, the third-generation Aixin Tongyuan hybrid precision NPU further enhances the support of operator types and different precision hybrid computing, optimizes the internal memory scheduling mechanism, and can efficiently support CNN and Transformer network models. The tool chain developed for Aixin Tongyuan 3.0 NPU, Pulsar2, further improves the functionality and usability of the tool chain, supports model deployment of mainstream deep learning training frameworks, supports two different quantization operations, PTQ and QAT, and meets the quantization and optimization functions of various scenarios.
The Transformer model is very simple and efficient to deploy on the Aixin Yuanzhi NPU. Taking the SwinT model as an example, the ONNX model of SwinT can be directly downloaded from the official ModelZoo of PyTorch. Without any model or operator modifications, SwinT can be quantitatively compressed and compiled, achieving a high energy efficiency ratio of 199 FPS/W. The AX650N, which was awarded the "Most Innovative Value Product Award" at the conference, is a high-performance chip product launched by Aixin Yuanzhi this year. Based on the third-generation NPU, the AX650N deploys Transformers on the edge and end sides with high performance, high precision, easy deployment, and low power consumption, making it the industry‘s leading Transformer landing platform.
Creating an ecological chip developer ecosystem with Aixin Pi Pro to assist in the landing of visual models
At the conference, Aixin Yuanzhi officially launched the developer kit - "Aixin Pai Pro", which is jointly created by hardware ecological partners. The developer kit aims to provide community developers with low-cost local experience of convenient deployment of visual models on the edge and end sides. The Aixinpai Pro is equipped with the AX650N and has high computing power and strong encoding and decoding capabilities, which can meet the industry‘s demand for high-performance edge intelligent computing. It achieves applications such as video structuring, behavior analysis, and state detection, and efficiently supports CNN models and Transformer visual models.
At the same time as the release of Aixin Pi Pro, Aixin Yuanzhi will also release a community version of the new generation AI tool chain Pulsar2 that integrates model quantification, model compilation, and model deployment functions, as well as rich development documents, facilitating users to quickly conduct product prototype verification and secondary development, and helping users to play greater value in fields such as smart cities, smart transportation, smart education, and smart manufacturing.
At present, Aixin Yuanzhi has mass-produced more than four generations of visual perception chips applied in different industries, and adheres to vertical integration based on algorithms, chips, and products, providing full stack solutions for partners and helping customers achieve the rapid implementation of the latest technology. Facing the rapid development of the artificial intelligence industry, Aixin Yuanzhi will also improve chip performance, reduce costs, and make edge intelligence more inclusive through continuous technological innovation, ultimately achieving the mission of "inclusive AI creates a better life".
About Aixin Yuanzhi:
Aixin Yuanzhi was established in May 2019. As an artificial intelligence visual perception chip research and development and basic computing platform company, Aixin Yuanzhi has formed a fully functional team from chip design, research and development to production. The core members have participated in the design and production of more than 10 chips and have rich experience in product planning and product implementation.
At present, Aixin Yuanzhi has successfully developed and mass-produced four generations of multiple end side and edge side intelligent visual perception chips, all of which have low-power advantages and excellent image processing capabilities. With the self-developed core technology - Aixin Smart Eye AI-ISP and Aixin Tongyuan Hybrid Precision NPU, Aixin chip products can meet the product needs of customers in different scenarios. The related products and solutions have been successfully implemented in East, South, and North China, and are widely used in fields such as smart cities, intelligent transportation, smart homes, smart manufacturing, and smart consumption.
|
Disclaimer: This article is transferred from other platforms and does not represent the views and positions of this site. If there is any infringement or objection, please contact us to delete it. thank you! |
Business consulting
13823761625
Mail me